Consensus analysis. (a) Sensitivity and specificity at identifying 1,556 consensus transcripts from the pool of the following evidence-based tracks: RefSeq, Known Gene, Ensembl, Gencode, AceView, ECgene and ExonWalk. The sensitivity and specificity of all tracks at identifying these consensus models is plotted and listed in Table 2. (b)Closest neighbor consensus, evaluated by switching the track of reference. This figure shows the number of evidence-based models from CCDS, RefSeq, UCSC Known Genes, Gencode, or AceView, ExonWalk and Ensembl whose intron-exon structure is exactly matched by the 25 tracks. Tracks are arranged in decreasing order of averaged detection sensitivity, defined here as the sum of all evidence-based models from these seven reference tracks detected exactly.